Computed tomography radiation dose reduction: effect of different iterative reconstruction algorithms on image quality.

Abstract

We evaluated the effects of hybrid and model-based iterative reconstruction (IR) algorithms from different vendors at multiple radiation dose levels on image quality of chest phantom scans. A chest phantom was scanned on state-of-the-art computed tomography scanners from 4 vendors at 4 dose levels (4.1 mGy, 3.0 mGy, 1.9 mGy, and 0.8 mGy). All data were reconstructed with filtered back projection (FBP) and reduced-dose data also with IR (iDose4, Adaptive Iterative Dose Reduction 3D, Adaptive Statistical Iterative Reconstruction, Sinogram-Affirmed Iterative Reconstruction, prototype Iterative Model Reconstruction, and Veo). Computed tomography numbers and noise were measured in the spine and lungs. Signal-to-noise ratios (SNR) and contrast-to-noise ratios (CNR) were calculated and differences were analyzed with the Friedman test. For all vendors, radiation dose reduction with FBP resulted in significantly increased noise levels (≤148%) as well as decreased SNR (≤57%) and CNR (≤58%) (P < 0.001). Conversely, IR resulted in decreased noise levels (≤48%) as well as increased SNR (≤94%) and CNR (≤94%). The SNRs and CNRs of the model-based algorithms at 80% reduced dose were similar to reference-dose FBP. Hybrid IR algorithms have the potential to reduce radiation dose with 27% to 54% and model-based IR algorithms with up to 80%.